Determinants of learning in developing countries

The success since 1960 in expanding the quantity of education in most

developing countries has shifted attention to education quality, especially as

Table 4-4: Krueger‘s Reanalysis of Hanushek's (1997) Class Size Studies

Results Hanushek‘s weights Studies equally weighted Studies weighted by journal impact factor Regression- adjusted weights (1) (2) (3) (4)

Positive & stat. sig. (%) 14.8 25.5 34.5 33.5 Positive & stat. insig. (%) 26.7 27.1 21.2 27.3 Negative & stat. sig. (%) 13.4 10.3 6.9 8.0 Negative & stat. insig. (%) 25.3 23.1 25.4 21.5 Unknown sign & stat. insig.

(%)

19.9 14.0 12.0 9.6

Ratio positive to negative 1.07 1.57 1.72 2.06

p-value* 0.500 0.059 0.034 0.009

106

measured by student performance on academic tests. Glewwe and Kremer

(2005) opine that most empirical studies of the determinants of years of

schooling and learning in both developed and developing countries are

retrospective studies, based on data generated by ordinary (non-experimental)

variation across schools and households. Hence, both economists and other

social scientists have used retrospective data to investigate the impact of

school and teacher characteristics on learning. The authors claim that the most

significant recent retrospective studies of the determinants of learning in

developing countries since 1990s are: the research on Ghanaian middle

schools by Glewwe and Jacoby (1994); the study of Jamaican primary schools

by glue and others (1995); the investigation of grade 8 students in India by

Kingdon (1996); and the paper on Philippines primary schools by Tan and

others (1997). The study by Glewwe and Jacoby (1994) on Ghana have

examined student achievement in 1988-89, using scores on reading (English

and Mathematics) in Ghanaian middle schools (grades 7 to 10). Eighteen

schools and teacher variables were examined, but most estimated effects were

small and statistically insignificant. The only statistically significant teacher

variable was teaching experience, but its effect was indirect, in contrast, school

facilities had larger impacts (Glewwe and Kremer, 2005, p.30). A study by

Glewwe and others (1995) used Jamaican data collected in 1990 to examine

the performance of primary school students in reading (English) and

mathematics. More than 40 schools and teacher characteristics were

examined, including pedagogical processes and management structure. Most

variables had statistically insignificant effects. The school variables with

107

only), teacher training within the past 3 years (mathematics), routine academic

testing of students (reading and mathematics), and the use of textbooks in

class (reading). The size of these estimated impacts (in standard deviations of

the test score variable) was lower than those for Ghana (Glewwe and Kremer,2005,p.31). Kingdon‘s(1996) study of India is based on data collected in 1991. Tests in reading (Hindi and English) and mathematics were given to students in ―class 8‖ (grade 8). Kingdon examined five teacher variable (years of general education, years of teacher training, marks received on official

teacher exams, years of teaching experience, and salary) and three school

variables (Class size, hours per week of academic instruction, and an index of

17 physical characteristics). The teacher variable with significant effects were

teacher exam marks, which had significant positive impacts on both mathematics and reading scores, and teachers‘ year of education, which had a significantly positive impact on reading scores (Glewwe and

Kremer,2005,p.31).

Tan, Lane and Coustere (1997), using data from 1990 and 1991, investigate

the impact of school and teacher variables on the mathematics and reading

scores of 2,293 first graders in the Philippines. Of the teacher variables, the

score on the subject knowledge test in reading had a positive impact on students‘ reading scores: a one standard deviation increase in the teacher‘s score raised student learning by 0.12 standard deviations. The same is true for mathematics scores: a one standard deviation increase in the teacher‘s score raised student learning by 0.10 standard deviations. Turning to school

characteristics, the impact of textbooks was unstable for both subjects, in some

108

percent level was the lack of adequate furniture, which was associated with a

drop of -0.32 standard deviations in math and 0.29 standard deviations in

reading (Glewwe and Kremer, 2005, p.31).

In all four studies, most school and teacher variables were not significantly

different from zero, although this could reflect both low sample sizes (163

students in Ghana and 355 in Jamaica) and high correlation among many of

these variables. While each study did find that one or more teacher variable

had statistically significant impacts. They differed widely across the studies.

Similarly, three of the four studies finds significant impacts of physical inputs

(the exception being the Jamaica study), but again the specific inputs vary

across different studies. Part of this variation could reflect differences in the

variables available in the data, and part could reflect large socioeconomic

differences across countries but, whatever the reason for this variation, the

conclusion is that there is no general result regarding which teacher and school

variables raise learning in developing countries (Glewwe and Kremer, 2005,

p.32). The offshoot of the above discussion assumes that the estimated impact

of these four retrospective studies are accurate, but also provides many

reasons to worry about biases in such estimates. Perhaps the underlying

relationship that is more motivated teachers, principals, and parents were more

likely to keep the building in good repair. The inability to observe certain children and household characteristics such as the child‘s innate ability and parental tastes for education also leaves lingering doubts. Finally, it is likely that

schools variables are measured with a large amount of error-examples have

been presented in Tanzania (distance to schools) and the Philippines (books

109

are often statistically insignificant (Glewwe and Kremer, 2005, p. 34).

Nascimento (2008) quoting Hanushek and Hedges and Greenwald, on the

variation in output, says that findings often point in opposite directions, fuelling

endless controversies on whether ―there is not strong or consistent relationship between school resources and student performance‖ (Hanushek, 1997, p.148) or ―school resources are systematically [and sufficient] related to student achievement […] to be educationally important‖(Hedges and Greenwald, 1996, p.90). He sums up, ―indeed, the degree of influence of school resources on student achievement seems to vary widely depending on the sample taken, the level of aggregation of the data, and methodology used‖ (Nascimento, 2008, p.26).